Amplitude Review

Three percent of revenue came from customers using five or more Amplitude products in early 2024. That number hit 24% this quarter. The multi-product attach is the story Spenser Skates' team has been pitching since the direct listing — and Q1 numbers say it's working.

Q1 2026 revenue was $94M, up 17%. ARR reached $374M, 727 customers spend over $100K, and 67% of that revenue comes from enterprises above 1,000 employees. The Statsig customer-book deal adds roughly $16M ARR, deepening Experiment, Analytics, Session Replay, AI Agents, and Guides and Surveys cross-sell. Atlassian, NBCUniversal, and Square anchor the references.

But AMPL still trades around $11, roughly an eighth of the 2021 direct-listing open at $50. Mixpanel and PostHog keep undercutting the mid-market, and GA4 owns the free tier. Standardize where product, growth, and experimentation share one event graph.

Event-based depth is what Amplitude sells, and Mixpanel can't match the user-path granularity at the analyst's desk. Behavioral Cohorts and Causal Insights compound — the longer you instrument cleanly, the more your retention math improves.

The Digital Analytics Platform thesis stitches Command AI ($45M, October 2024), Iteratively (2021), Web Experimentation, and the June 2025 AI Agents into one suite. The Plus tier at $49/month and 300K MTUs gives PMs a real on-ramp; Growth and Enterprise carry the experimentation depth Heap and PostHog still don't.

But the catch is the agentic layer. The February 2026 AI Agents — Dashboard Monitoring Agent, Session Replay Agent — assume your event taxonomy is clean. If instrumentation drifted across three reorgs, the agents amplify noise. The 3-year bet is whether Amplitude's depth survives LLM-native query tools like Hex Magic and ThoughtSpot Sage eating the analyst job.

Amplitude prices the user. Mixpanel prices the event. Plus lists $49/month for 300K MTUs or 25M events, whichever you hit first. Mixpanel Growth runs $0 for the first 1M events, then $0.28 per 1K. Two unit primitives, two forecasts, and your invoice depends on which user pattern you ship.

Starter is generous: 10K MTUs, 2M events, Session Replay included. Plus jumps to $49/month but Session Replay flips to a paid add-on at that tier. Growth and Enterprise both hide pricing behind a sales call.

The catch is Anonymous Tracking. The docs indicate a device-ID visitor counts as an MTU before they ever log in, which inflates the line item nobody models. But the Startup Scholarship ships one free Growth year for under-$10M, under-20-employee teams, and Behavioral Cohorts justify the Plus sticker for a 50-person product org.

Iteratively defines events in a JSON schema before the SDK fires them — analysts catch typos in code review, not in Pathfinder a sprint later. That governance step is what Heap's autocapture model deliberately skips, and it shows up the first time someone renames a property.

Behavioral cohorts feed Audiences, and Audiences pushes to Braze with One-Time, Scheduled, or Real-Time sync — the marketer's loop closes without a CDP in the middle. Experiment runs feature flags and A/B tests off the same identity graph the analytics charts use, however the flagging UI still trails LaunchDarkly on guardrails.

Plus at $49/month caps at 300K MTUs or 25M events, so mid-growth teams hit a Growth-tier sales call earlier than Mixpanel's published plan forces. The tradeoff is governance and depth versus PostHog's self-host story or Heap's instrumentation-light approach.

The Ampli wrapper does the thing analytics tools usually leave to copy-paste. You define events in the tracking plan, run ampli pull, and the SDK ships type-checked methods that match the spec — JavaScript, iOS, Android, React Native, Unity, Node, the lot. That alone saves a week of taxonomy arguments.

The Starter tier is free up to 10K MTUs, Plus jumps to $49 a month for 300K. Mixpanel's free tier counts events instead of users, which feels cleaner for high-volume anonymous traffic. The Behavioral Graph auto-detects journeys and the Cohort API exports audiences cleanly — but the SQL Editor is read-only, so anything fancy round-trips through Notebooks or a real warehouse.

Three months in, the cost is the MTU meter. Anonymous device IDs count too, which is fine until a marketing campaign doubles your traffic and the bill follows. Beautiful taxonomy. Strict meter.

The market priced this for terminal decline at $6.19, down 89% from the $80 IPO peak. Q1 2026 didn't deliver that — $93.5M revenue up 17%, NRR back to 106% from 100% a year ago, $427M RPO up 31%. Multiproduct attach is the story management keeps selling and the numbers now back it.

PostHog is the real threat — $57.5M ARR, roughly 99% YoY growth, open-core, 108K company installs eating SMB from below. Mixpanel still cheaper. GA4 still free. AI Agents and the MCP integration with Claude and Cursor read as actual enterprise plays, not panic ships.

The catch: 17% at $374M ARR is decent, not great, and the market won't re-rate until four straight quarters land. Iteratively for CDP was strategic but underbaked. Event data exports clean — exit is fine. Long-term viability is the open bet.